Taking Profits: What Actually Works## A 10-Year Backtest of Systematic Profit-Taking Strategies¶
Summary of FindingsBefore we dive in, here's where this research ends up:The Simple Answer: It depends entirely on what you own and how you reinvest.- For outlier-heavy portfolios (think early NVDA positions): Trimming costs you millions. Buy-and-hold crushed every trimming strategy by 20-50%.- For index-focused portfolios (60% ETFs, 40% stocks): The right trimming strategy beat buy-and-hold by 52% while cutting max drawdown by 18%.- The critical variable: Not your trim threshold, but when and how you reinvest. Market-timing reinvestment destroyed returns. Gradual reinvestment during volatility created alpha.This isn't a story about finding "the best strategy." It's about understanding how systematic rules interact with portfolio composition and market conditions. Let's walk through how I figured this out.¶
The Question That Started EverythingI trimmed my NVDA position at +50% in 2023. Sold at $140. It hit $950 (split-adjusted). I lost over $100,000 in opportunity cost because I "took profits responsibly."That experience haunted me. Was systematic trimming—selling 20% when positions gain 50%, 100%, 150%—just a way to feel smart while bleeding returns? Or does it actually work if you do it right?I needed data. Not opinions, not anecdotes. Ten years of real market data testing every variation I could think of.¶
Methodology### The Framework- Period: 2015-2024 (10 years, 2,477 trading days)- Initial Capital: $100,000- Data Source: Yahoo Finance historical prices- Metrics: CAGR, Sharpe ratio, Sortino ratio, max drawdown, volatility- Test Count: 42 distinct strategies across 3 portfolio configurations### Three Phases of TestingPhase 1: Equal-weight 6-stock portfolio (AAPL, MSFT, NVDA, TSLA, SPY, QQQ)—testing core trimming mechanics.Phase 2: Added volatility-based trimming and gradual reinvestment modes—testing if smarter triggers and timing could improve results.Phase 3: Index-focused portfolio (60% ETFs, 40% stocks)—testing realistic investor allocations.### Strategy VariationsTrimming Types (5 approaches):- Fixed thresholds: Trim 20% at +50%, +100%, +150% gains- Momentum-guided: Dynamic thresholds (1.5× to 2.5× rolling average)- Volatility-based: Trim more when volatility spikes (1.5× to 2.5× thresholds)Reinvestment Modes (6 approaches):- Immediate: Instant reinvestment (pro-rata or SPY)- Gradual (DRIP): 25% per week over 4 weeks- Market-timing: Wait for 5% S&P dip- Yield/Volatility-based: Gradual reentry using dividend yield and VIX signals### Key AssumptionTransaction costs and taxes defaulted to 0% for baseline analysis. The backtest engine supports toggling both (0.05-0.5% per transaction, 15-37% capital gains tax), documented in /docs/COST_TAX_MODELING.md. Adding realistic costs (0.1% transaction + 20% tax) reduces all strategies proportionally but preserves relative rankings.¶
Phase 1: The NVDA Trap### Portfolio ConfigurationEqual-weight allocation: $16,667 per ticker (AAPL, MSFT, NVDA, TSLA, SPY, QQQ).### What Happened¶
Buy-and-hold crushed every trimming strategy. Final value: $5.4M (50.1% CAGR). Best trimming strategy: $4.3M (46.7% CAGR). Worst: $1.1M (28.1% CAGR).That's not a rounding error. That's the difference between generational wealth and a nice car.### The Root Cause: NVDANVDA gained **28,057%** over the period. It went from $0.48 to $136.04.Every time we trimmed at +50%, we sold at $1. At +100%, we sold at $2. At +150%, we sold at $5. Meanwhile, it went to $136. Trimming systematically cut exposure to the decade's biggest winner at precisely the wrong time.Lesson 1: In portfolios with lottery-ticket outliers, ANY profit-taking is catastrophically expensive.
Phase 2: The Innovation That Failed### The Dip-Buy HypothesisWhat if we didn't reinvest immediately? What if we waited for market drops and bought SPY/QQQ on 5% dips?This felt clever. Take profits at peaks, redeploy at troughs. Buy low, sell high. Textbook.### ResultsThe dip-buy strategy executed 6-9 successful dip purchases over 10 years. And it still underperformed immediate reinvestment.- Buy-and-hold: $5.4M (50.1% CAGR)- **Best dip-buy strategy**: $2.7M (39.8% CAGR)- Immediate SPY reinvestment: $3.1M (41.2% CAGR)### Why It FailedOpportunity cost. Cash waiting for dips sat idle during the longest bull market in history. The benefit of lower entry prices couldn't overcome the cost of missing months of compounding.In sideways or bear markets, this might work. In a raging bull market from 2015-2021, every day in cash was a day of lost returns.Lesson 2: Market timing—even "smart" timing—costs you in strong bull runs.¶
Phase 3: The Portfolio Composition Breakthrough### The RealizationNobody buys NVDA at $0.48. That's lottery-level luck. Real investors hold mostly index funds with some individual stock positions.I rebuilt the portfolio to reflect actual investor behavior:- 60% index funds: SPY 30%, QQQ 20%, VOO 10%- 40% individual stocks: AAPL 15%, MSFT 15%, TSLA 10%This is an illustrative allocation, not an optimized one. It represents index-heavy portfolios common among retail investors.### Results: Everything Flipped¶
- Buy-and-hold: $689k (21.7% CAGR, -46.3% max drawdown, 0.90 Sharpe)- **Best trimming (Volatility-2.5× pro-rata)**: $1,047k (26.98% CAGR, -37.8% max drawdown, 1.12 Sharpe)The best trimming strategy didn't just match buy-and-hold. It beat it by 52% with an 18% lower max drawdown.This completely inverts the Phase 1 conclusion.### What Changed?Reduced outlier exposure. With 60% in diversified ETFs, no single stock could dominate outcomes like NVDA did in Phase 1.Volatility became signal. In index-heavy portfolios, volatility spikes often precede mean-reversion. Trimming during volatility captured gains before pullbacks.Gradual reinvestment worked. Spreading reentry over weeks during volatile periods created dollar-cost averaging benefits that immediate reinvestment missed.Lesson 3: Portfolio composition determines whether trimming works. Strategy mechanics matter less than what you own.
The Reinvestment Mode Discovery¶
The gap between strategies wasn't about when you trimmed. It was about how you reinvested.### Immediate Reinvestment- Pro-rata: Maintained exposure to high-growth stocks. Strongest performer in Phase 1.- SPY: Rotated profits to slower index. Underperformed by moving capital from winners to laggards.### Gradual Reinvestment (DRIP: 25%/week)- Pro-rata DRIP: Beat immediate pro-rata by 3-5% in volatile periods.- SPY DRIP: Still underperformed, but gap narrowed vs immediate SPY.### Market-Timing Reinvestment- Dip-buy (5% S&P drop): Worst performer. Opportunity cost destroyed returns.### Yield/Volatility-Based Reinvestment- Gradual reentry using dividend yield + VIX: Best overall. Combined dollar-cost averaging with volatility signals.- Conceptually similar to volatility harvesting: buying when variance is high, capturing mean-reversion.
Key Insight: Gradual reinvestment during volatile periods creates alpha. This isn't market timing (predicting direction)—it's volatility timing (spreading entries when prices swing). The former fails; the latter works.Lesson 4: Don't just take profits. Structure how you redeploy them.
Volatility-Based Trimming: When Algorithms Fail### The HypothesisWhat if we trimmed based on volatility instead of fixed price thresholds? Trim more when stocks are whipsawing, less when they're stable.This is algorithmic loss aversion: avoid holding positions during high-variance regimes.### ResultsVolatility-based trimming with immediate reinvestment underperformed fixed thresholds in Phase 1 (outlier-heavy portfolio). It over-trimmed NVDA during its explosive growth periods, which happened to be high-volatility.But in Phase 3 (index-focused portfolio), volatility-2.5× with gradual pro-rata reinvestment produced the highest CAGR (26.98%).### Why?In concentrated portfolios: High volatility often precedes explosive growth (NVDA 2020-2024). Trimming during volatility cuts winners early.In diversified portfolios: High volatility often precedes mean-reversion. Trimming captures gains before pullbacks, then gradual reentry buys dips.This is a behavioral finance lesson: The same algorithm produces opposite outcomes depending on portfolio structure. There's no universal "smart" trigger.Lesson 5: Algorithmic rules interact with portfolio composition. Test your strategy on your holdings, not generic backtests.¶
Rolling Performance Analysis¶
Looking at 1-year rolling returns reveals when trimming helped vs hurt:- 2016-2019 (steady bull market): Trimming lagged buy-and-hold. Opportunity cost from reducing winners.- 2020 (COVID crash/recovery): Trimming with gradual reinvestment outperformed. Captured pre-crash gains, bought recovery dips.- 2021-2022 (volatile chop): Trimming matched or beat buy-and-hold. Profit-taking during peaks, reentry during dips.- 2023-2024 (AI boom): Trimming lagged again. Missing runaway momentum.Pattern: Trimming works in volatile/choppy markets. Buy-and-hold works in steady trends.Lesson 6: No strategy dominates all regimes. Trimming is volatility insurance, not a return maximizer.
Sensitivity Analysis¶
Testing trim thresholds (50% to 200%) against trim sizes (10% to 50%) reveals:- Higher thresholds + smaller trim sizes: Fewest regrets. Let winners run, take modest profits.- Lower thresholds + larger trim sizes: Maximum regret. Over-trim early, miss compounding.- Sweet spot: 100-150% threshold, 15-20% trim size.But this is portfolio-specific. In Phase 1 (NVDA-heavy), no trimming beats buy-and-hold. In Phase 3 (index-heavy), strategic trimming wins.Lesson 7: Optimize trim parameters for your portfolio, not generic rules.
Drawdown Analysis¶
Maximum drawdown comparison:- Buy-and-hold: -46.3% (March 2020 COVID crash)- Best trimming: -37.8% (Volatility-2.5× pro-rata DRIP)That 8.5% difference is the psychological benefit of trimming. Smoother ride, less panic-selling risk.During the COVID crash, trimming strategies had already taken some profits in Feb 2020 (volatility spike). More cash cushion when markets tanked. Gradual reinvestment in March-April bought near the bottom.Lesson 8: Trimming isn't just about returns. It's about surviving drawdowns without panic-selling.